International Journal of Engineering & Technology, 7 (2.

31) (2018) 76-79

International Journal of Engineering & Technology

Website: www.sciencepubco.com/index.php/IJET

Research paper

Handover forecasting in 5G using machine learning

A. Suresh Kumar1*, S. Vanmathi2, B. Praveen Sanjay3, S. Ramya Bharathi4, M. Sakthi Meena5
1Assistant
Professor, Department of Information Technology, Sri Krishna College of Technology, Coimbatore, Tamil Nadu, India.
2UG Student, Department of Information Technology, Sri Krishna College of Technology, Coimbatore, Tamil Nadu, India.
3UG Student, Department of Information Technology, Sri Krishna College of Technology, Coimbatore, Tamil Nadu, India.
4UG Student, Department of Information Technology, Sri Krishna College of Technology, Coimbatore, Tamil Nadu, India.
5UG Student, Department of Information Technology, Sri Krishna College of Technology, Coimbatore, Tamil Nadu, India.

Abstract

Communication plays a major role in human’s life. Without network communication can’t be done, to achieve proper and effective
communication different generations of networks are introduced. Each generation has its own features and perspective of
communication, but till now there is no network properly makes people to communicate. Many researches says 5G network will rule the
network world as it satisfies all the effective network goals. This paper is proposed to obtain all the goals of a communication network by
making proper handover with the help of machine learning.
Here we have used two main algorithms to make our 5G handover process by clustering and classifying. Clustering is a process of
making the datasets into single units of every users and classification is a process of classifying user’s clustered datasets into common
path using prediction and forecasting. For clustering we are using K-means and for classification we are using Random Forest algorithm.
By using the algorithms the datasets which is being predicted and forecasted is stored in the cloud. Here cloud technology is used as a
platform for developing datasets associated with internet. 5G network adapts to any form technology easier and here we have used all the
essential technologies under machine learning. This paper deals with all the above methodologies effectively with newer combinations of
algorithms along with proper solutions.

1. Introduction expands the bandwidth capacity of this system. Mainly higher

bandwidth systems have more users’ occupancy.
5G network support MIMO for improving energy efficiency by
Communication plays an important role in everyone life’s
radiating which gives 100 times better capacity system than the
.Networking is a construction design and uses of a network which
other generation networks. It also uses spectral sharing where the
includes the physical (cabling, hub, bridge, router, and so
target broadband network can be easy identify and gives a
forth),the selection and uses telecommunication protocol by
centralized solution for all the mapping of networks centralized
managing the networks .It also gives the establishment of
sharing spectrum techniques is used to map the database using
operation policies and procedures related to the network [1]. 5G
geo- location ,were the database of the users can be stored and
networking is a wireless board band technology which supports all
retrieved anywhere in the world. In this paper we have proposed a
formats of networking devices. An important goal of 5G is to
system which uses all the 5G properties with the help of machine
erase the differences between wire line and wireless networking .It
learning. Handover using machine learning gives more quality of
will provide better speed and coverage than the current 4G by
services by mapping to the required target services. By using
providing qualified and better handover.5G network incorporates
machine learning we can predict and forecast the user’s path while
all newer technologies with higher range of frequency ,proper
establishing handover [3].and it can store huge amount of data
antenna configuration along with deployment of ultra-dense
under a single handover service. In this paper we will consider
device to device communication [2]. Handover is a bit difficult
three main aims along with machine learning 1. Quality of service
process even though lots of newer technologies were introduced
2.mobility 3.bandwidth.Quality of services is used to refer the
.Handover consist of two form 1.vertical Handover 2.Horizontal
ability of a network to provide improved service to select network
Handover. Vertical Handover refers to a networking changing the
technologies. Quality of service is a single network element,
type of connection to access a supporting infrastructure. It refers
which includes queuing, scheduling, and traffic shaping features.
to automatic switching to communicate from one to other
In this paper QoS mainly focuses on user’s side by giving required
technology. It enables the exploitation of higher bandwidth and
handover of specified bandwidth capacity. Mobility of services is
lower for networks like local area networks. It also provides
a concept that aims to combine all modes of transport. A quality or
extended coverage for cellular networks. The vertical handover
capability which permits them to move from place to place while
process imposes important technical challenges. Horizontal
retaining the ability to fulfill their primary mission .These devices
handover a network always occurs in same radio access
provides several benefits to users and in this paper mobility plays
technology handover process occurs in different networks.
an important role as 5G network always finds a better solution for
Horizontal handover is a mobile terminal changes its point of
this problem. While the user is in handover proper handover
connection within the same type of networks. In this paper we are
switch is made by using two algorithms namely K-means and
using vertical handover as it uses automatic mapping of the target
Random F o r e s t algorithm. The bandwidth is defined as a link
handover channel. Once the required handover is given definitely
for communication [4] higher bandwidth capabilities will make a
the network becomes more effective and qualified. Vertical
better system.
handover helps in giving handover to different systems were it
Machine learning is an effective newer technology which makes
Copyright © 2018 Authors. This is an open access article distributed under the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
International Journal of Engineering & Technology 77

handover using automatic prediction and forecasting. It is a machine learning which can address their management challenges.
relatively new discipline within computer science that provides The author’s expectation in this paper is 5G network can bring
collection of data techniques. Machine learning is an international more intelligent level in monitoring, managing networks and
forum for research on computational approaches. Here different applications. UdiMargolin and Alberto Mozo[8] gives some
algorithms were applied for different purpose like clustering and technique to identify and solve the noisy neighbor in NFV
forecasting application. In this paper we will consider two main infrastructure in 5G. They used simple and standard automated
types of algorithm. 1. k means, 2.Random Forest. K-means model to detect the noisy neighbor problem with an accuracy of
clustering is an algorithm used to cluster the dataset into K more than 90%.
clusters. Every cluster of k has a center point known as centroid,
by using the center point it clusters all the related dataset into a 3. Proposed system
single-single unit. In this paper we have used K-means for
clustering different data types into single unit of related data.
Random forest algorithm is used for classification of the clustered 5G networks propose the next generation communication
dataset; it makes the dataset into different tree like structures. As standards. 5G is an improved network which will be used in the
more number of trees will increase the accuracy of the system. future. Next generation wireless networks are expected to support
The tree like structure is formed by using prediction of the dataset extremely high data rates. Future smart 5G terminals expected to
path by using training dataset. Machine learning system develops access the bandwidth which will impact entire mobile networks
automatically and customizes themselves to the individual users. and it will transformed into 5G using technologies like NFV,SDN,
In this paper we uses cloud to store datasets were it can be enabling speed, agility and efficiency. Handover is a process
retrieved, updated, added and removed easily. which transfers upcoming data or ongoing data from one channel
network to next channel. When a mobile terminal moves outside
the coverage area of the base station it takes appropriate measures
2. Related works assumed by the network management. The quality of bandwidth is
required to ensure the base station.
Network handover is a tedious process in communication, for Mobility and handover are based on services, allows the user to
overcoming the difficulties in handover many researches using move from one cell range to another or to switch to nearest cell for
different generations of networks is done. This paper deals with performance. The three main parameters considered in this paper
5G networking for establishing an effective handover. The are quality of services, bandwidth, and mobility. To increase these
following researches were made with the aim of improvised and parameters by using machine learning, two algorithms are used.
effective handover implementation. The authors David Soldani One is k-means for clustering and other is random forest algorithm
and Antonio Manzalini[1] gives a solution for 5G infrastructure for classification. The upcoming flow diagram will illustrated all
which will be converted into a 3D collaborative and immersive the essential things to be done for constructing the system. In
service. Here they use ultra low latency nerves system for machine learning, Clustering is grouping of a set of observation
supporting 3D. Authors Liang Gong, Bo Rong and KejieLu[2] into subsets called cluster. The unsupervised procedure is known
proposed a system which uses software defined network(SDN) as clustering, which involves grouping of data into categories
which is based on intelligent model in heterogeneous network. based on similarity, example grouping of all calls, texts, videos,
This paper is made to improve the system capacity and to achieve audio, and media from every users. In this paper we have used K-
users demand dynamically. As a result of this system the network means to cluster different data types into singe- single unit of
becomes scalable, reliable and positively implementable using datasets.
SDN infrastructure.
In [3] this paper the author’s uses call handling mechanism,
conventional handling mechanism and different handover
prioritization schemes for improving 5G handover. They use
balancing scheme and cell overlap scheme for reducing traffic and
system complexity while giving a handover. They also uses
mathematical formulas for obtaining a better 5G system. Michele
polese and Marco Giordani [4] are the authors proposed a
connectivity of dual systems from 4G to 5G .First comprehensive
end – to-end evaluation of handover mechanism in mm wave
cellular systems was provided by this paper. It contains complete
details of medium access control, radio link control and transport
protocols. In [5] this paper the authors Marco Mezzavilla and
Michele Zorzi use high directionality techniques to improve signal
quality and extend area coverage. This paper proposed integrated
LTE network architecture and mm wave radio access technologies
to provide ultra-reliable services to mobile users and it provides
alternatives for integration and shows how simulation can be used
to assess and compare their performance. This[6]paper authors Le
LuongVy and Li-Ping Tung told that handover is the important
key for improving network quality and mobility performance in
mobile networks. They implies that using of macro cells, small
cells and ultra-dense networks make handover more complicated.
Authors used more than 6000 cells of real networks for
successfully identifying the separating handover behaviours,
forecasting future number of handover attempts and detecting
abnormal handover behaviours of cells.
In [7] this paper authors Teodora Sandra Buda and Haytham
Assem explained about the limitations of current network and
service management and completely explains the challenges that
5G faces from management perspective. The crisp thing of this
Figure 1: Flow diagram of ML system
paper is it presents number of use cases and scenario of 5G in
78 International Journal of Engineering & Technology

K-means uses three main steps 1. K centroid initialization related data were clustered into a single unit. Here blue colour
2.finding nearest centroid and clustering them into unit 3.using represents centroid for global system and remaining were data.
arithmetic notations finding newer closest clusters, finally repeat Locally each dataset has centroid and does clustering function of
step 2 and 3 for making newer and effective clusters. The centroid different data types.
initialization is making centre of node which covers all the raw
dataset around them. Once centroid is fixed the nearest dataset is
founded out and marked. After marking by using notation the
datasets were clustered. Once clustering is completed the clustered
dataset is sent to the training face for prediction and forecasting.
The prediction and forecasting is done using Random Forest
algorithm. Dataset training plays an important role for prediction
and forecasting. When the phone is moving away from the area
covered by one cell and entering into the area covered by another
cell and the call is transferred to second cell to avoid call
termination when the phone gets outside the range of first cell. In
order to avoid the interference, the cell is transferred to a different
channel in the same cell or to different channel in another cell
when the channel used by the phone becomes interfered. To
reduce the interference to smaller neighboring cell due to near-far
effect when phone still has an excellent connection to its current
cell. When a fast travelling user stops then the call may be
transferred to smaller macro cell for fast travelling users, this
works in reverse to reduce the potential interference to other cells
or users. Figure 3: Random forest classifications 1
Random forest algorithm is used to do all above mentioned things
with high efficiency. This algorithm makes tree like structure by Figure represents that data after training is classified based on
using the training dataset which searches and finds the next prediction in a circular tree way.
handover target easily. In this paper this algorithm is mainly used
to find the prediction path and to makes the dataset moved in the
right way. The dataset training is made using the regular watching
of the dataset path. Once the dataset is trained under a specified
path all the nearest network which comes under path is allocated
to the dataset users. All the trained datasets will fall under their
own trained path, when the user from one trained path wants to get
joined in another trained dataset due to change in path. In that
situation forest algorithm pays way to the newer dataset to be
added in the required trained dataset. For each and every user
proper handover is fixed by seeing their path to the nearest
network by prediction. By mapping to the required bandwidth
network service quality of service is more. This is the way were
we provide an effective handover using K-means and Random
Forest algorithm under machine learning.

4. Experiment and result Figure 4: random forest classifications 2

The dataset is proposed into clustering and classification phases
and finally the required target handover is switched. Here we are
using matlab for implementing the two phases into an effective
processing. First clustering is done them classification is made.
Both these processes use user datasets.
The figure gives classification in linear order were dataset gets
changed to another path from one predicted set to another set. The
path changing does not affect the handover process because the
changing dataset can go the required predicted path once it is
changed

5. Conclusion and future enhancement

5G ML handover system using machine learning is an effective
and qualified system to improved handover in 5G network. Here
we have used more accurate algorithms to achieve handover
without traffic and interception. Cloud storage makes this system
even more effective for online usages. By using prediction and
forecasting the handover in this system is easier than the other
generation networks. This system can be enhanced by adding
upcoming technology. This system can be also developed by using
other core related algorithms such mm wave format and MIMO.
Many researches were made to implement this formats into
existence by using machine learning.

References
Figure 2: K-means clustering
[1] http://www.future.bs/network.htm
Figure 2 shows that data is clustered using centroid and all the [2] The differences between 2.4GHz and 5GHzWireles
International Journal of Engineering & Technology 79

https://www.tplink.com/in/faq-499.html
[3] Use Machine Learning to Predict the Quality
https://medium.freecodecamp.org/using-machine-learning-to-
predict-the- quality-of-wines-9e2e13d7480d
[4] Chakraborty T, “EC3: Combining Clustering and Classification for
Ensemble Learning”, IEEE International Conference on Data
Mining (ICDM), (2017), pp.781-786
[5] Soldani D & Manzalini A, “Horizon 2020 and beyond: on the 5G
operating system for a true digital society”, IEEE Vehicular
Technology Magazine, Vol.10, No.1,(2015), pp.32-42.
[6] Sun S, Gong L, Rong B & Lu K, “An intelligent SDN framework
for 5G heterogeneous networks”, IEEE Communications
Magazine, Vol.53, No.11,(2015), pp.142-147.
[7] Khan J, Abbas A & Khan K, “Cellular Handover approaches in 2.5
G to 5G Technology”, International Journal of Computer
Applications, Vol.21, No.2,(2011), pp.0975-8887.
[8] Polese M, Giordani M, Mezzavilla M, Rangan S & Zorzi M,
“Improved handover through dual connectivity in 5G mmWave
mobile networks”, IEEE Journal on Selected Areas in
Communications, Vol.35, No.9,(2017), pp.2069-2084.
[9] Giordani M, Mezzavilla M & Zorzi M, “Initial access in 5G
mmWave cellular networks”, IEEE Communications
Magazine, Vol.54, No.11,(2016), pp.40-47.
[10] Tung LP & Lin BSP, “Big data and machine learning driven
handover management and forecasting”, IEEE Conference on
Standards for Communications and Networking (CSCN), (2017),
pp. 214-219.
[11] Redana S, Kaloxylos A, Galis A, Rost P & Jungnickel V, “View on
5G Architecture”, White paper of the 5G-PPP architecture WG,
(2016).
[12] Margolin U, Mozo A, Ordozgoiti B, Raz D, Rosensweig E & Segall
I, “Using Machine Learning to Detect Noisy Neighbors in 5G
Networks”, arXiv preprint arXiv:1610.07419, (2016).